Offsite-tuning is a privacy-preserving method for tuning large language models (LLMs) by sharing a lossy compressed emulator from the LLM owners with data owners for downstream task tuning. This approach protects the privacy of both the model and data owners. However, current offsite tuning methods often suffer from adaptation degradation, high computational costs, and limited protection strength due to uniformly dropping LLM layers or relying on expensive knowledge distillation. To address these issues, we propose ScaleOT, a novel privacy-utility-scalable offsite-tuning framework that effectively balances privacy and utility. ScaleOT introduces a novel layerwise lossy compression algorithm that uses reinforcement learning to obtain the importance of each layer. It employs lightweight networks, termed harmonizers, to replace the raw LLM layers. By combining important original LLM layers and harmonizers in different ratios, ScaleOT generates emulators tailored for optimal performance with various model scales for enhanced privacy protection. Additionally, we present a rank reduction method to further compress the original LLM layers, significantly enhancing privacy with negligible impact on utility. Comprehensive experiments show that ScaleOT can achieve nearly lossless offsite tuning performance compared with full fine-tuning while obtaining better model privacy.

In this work, we provide a systematic analysis of how large language models (LLMs) contribute to solving planning problems. In particular, we examine how LLMs perform when they are used as problem solver, solution verifier, and heuristic guidance to improve intermediate solutions. Our analysis reveals that although it is difficult for LLMs to generate correct plans out-of-the-box, LLMs are much better at providing feedback signals to intermediate/incomplete solutions in the form of comparative heuristic functions. This evaluation framework provides insights into how future work may design better LLM-based tree-search algorithms to solve diverse planning and reasoning problems. We also propose a novel benchmark to evaluate LLM's ability to learn user preferences on the fly, which has wide applications in practical settings.

The discovery of customer intention from dialogue plays an important role in automated support system. However, traditional text clustering methods are poorly aligned with human perceptions due to the shift from embedding distance to semantic distance, and existing quantitative metrics for text clustering may not accurately reflect the true quality of intent clusters. In this paper, we leverage the superior language understanding capabilities of Large Language Models (LLMs) for designing better-calibrated intent clustering algorithms. We first establish the foundation by verifying the robustness of fine-tuned LLM utility in semantic coherence evaluation and cluster naming, resulting in an accuracy of 97.50% and 94.40%, respectively, when compared to the human-labeled ground truth. Then, we propose an iterative clustering algorithm that facilitates cluster-level refinement and the continuous discovery of high-quality intent clusters. Furthermore, we present several LLM-in-the-loop semi-supervised clustering techniques tailored for intent discovery from customer service dialogue. Experiments on a large-scale industrial dataset comprising 1,507 intent clusters demonstrate the effectiveness of the proposed techniques. The methods outperformed existing counterparts, achieving 6.25% improvement in quantitative metrics and 12% enhancement in application-level performance when constructing an intent classifier.

Traditional evaluation metrics like BLEU and ROUGE fall short when capturing the nuanced qualities of generated text, particularly when there is no single ground truth. In this paper, we explore the potential of Large Language Models (LLMs), specifically Google Gemini 1, to serve as automatic evaluators for non-standardized metrics in summarization and dialog-based tasks. We conduct experiments across multiple prompting strategies to examine how LLMs fare as quality evaluators when compared with human judgments on the SummEval and USR datasets, asking the model to generate both a score as well as a justification for the score. Furthermore, we explore the robustness of the LLM evaluator by using perturbed inputs. Our findings suggest that while LLMs show promise, their alignment with human evaluators is limited, they are not robust against perturbations and significant improvements are required for their standalone use as reliable evaluators for subjective metrics.

The integration of artificial intelligence (AI) into the workplace is advancing rapidly, necessitating robust metrics to evaluate its tangible impact on the labour market. Existing measures of AI occupational exposure largely focus on AI's theoretical potential to substitute or complement human labour on the basis of technical feasibility, providing limited insight into actual adoption and offering inadequate guidance for policymakers. To address this gap, we introduce the AI Startup Exposure (AISE) index-a novel metric based on occupational descriptions from O*NET and AI applications developed by startups funded by the Y Combinator accelerator. Our findings indicate that while high-skilled professions are theoretically highly exposed according to conventional metrics, they are heterogeneously targeted by startups. Roles involving routine organizational tasks-such as data analysis and office management-display significant exposure, while occupations involving tasks that are less amenable to AI automation due to ethical or high-stakes, more than feasibility, considerations -- such as judges or surgeons -- present lower AISE scores. By focusing on venture-backed AI applications, our approach offers a nuanced perspective on how AI is reshaping the labour market. It challenges the conventional assumption that high-skilled jobs uniformly face high AI risks, highlighting instead the role of today's AI players' societal desirability-driven and market-oriented choices as critical determinants of AI exposure. Contrary to fears of widespread job displacement, our findings suggest that AI adoption will be gradual and shaped by social factors as much as by the technical feasibility of AI applications. This framework provides a dynamic, forward-looking tool for policymakers and stakeholders to monitor AI's evolving impact and navigate the changing labour landscape.

Recently, LoRA has emerged as a crucial technique for fine-tuning large pre-trained models, yet its performance in multi-task learning scenarios often falls short. In contrast, the MoE architecture presents a natural solution to this issue. However, it introduces challenges such as mutual interference of data across multiple domains and knowledge forgetting of various tasks. Additionally, MoE significantly increases the number of parameters, posing a computational cost challenge. Therefore, in this paper, we propose MoSLD, a mixture-of-shared-LoRAs model with a dropout strategy. MoSLD addresses these challenges by sharing the upper projection matrix in LoRA among different experts, encouraging the model to learn general knowledge across tasks, while still allowing the lower projection matrix to focus on the unique features of each task. The application of dropout alleviates the imbalanced update of parameter matrix and mitigates parameter overfitting in LoRA. Extensive experiments demonstrate that our model exhibits excellent performance in both single-task and multi-task scenarios, with robust out-of-domain generalization capabilities.

Powerful large language models (LLMs) are increasingly expected to be deployed with lower computational costs, enabling their capabilities on resource-constrained devices. Post-training quantization (PTQ) has emerged as a star approach to achieve this ambition, with best methods compressing weights to less than 2 bit on average. In this paper, we propose Channel-Relaxed Vector Quantization (CRVQ), a novel technique that significantly improves the performance of PTQ baselines at the cost of only minimal additional bits. This state-of-the-art extreme compression method achieves its results through two key innovations: (1) carefully selecting and reordering a very small subset of critical weight channels, and (2) leveraging multiple codebooks to relax the constraint of critical channels. With our method, we demonstrate a 38.9% improvement over the current strongest sub-2-bit PTQ baseline, enabling nearer lossless 1-bit compression. Furthermore, our approach offers flexible customization of quantization bit-width and performance, providing a wider range of deployment options for diverse hardware platforms.

Creating AI systems that can interact with environments over long periods, similar to human cognition, has been a longstanding research goal. Recent advancements in multimodal large language models (MLLMs) have made significant strides in open-world understanding. However, the challenge of continuous and simultaneous streaming perception, memory, and reasoning remains largely unexplored. Current MLLMs are constrained by their sequence-to-sequence architecture, which limits their ability to process inputs and generate responses simultaneously, akin to being unable to think while perceiving. Furthermore, relying on long contexts to store historical data is impractical for long-term interactions, as retaining all information becomes costly and inefficient. Therefore, rather than relying on a single foundation model to perform all functions, this project draws inspiration from the concept of the Specialized Generalist AI and introduces disentangled streaming perception, reasoning, and memory mechanisms, enabling real-time interaction with streaming video and audio input. The proposed framework InternLM-XComposer2.5-OmniLive (IXC2.5-OL) consists of three key modules: (1) Streaming Perception Module: Processes multimodal information in real-time, storing key details in memory and triggering reasoning in response to user queries. (2) Multi-modal Long Memory Module: Integrates short-term and long-term memory, compressing short-term memories into long-term ones for efficient retrieval and improved accuracy. (3) Reasoning Module: Responds to queries and executes reasoning tasks, coordinating with the perception and memory modules. This project simulates human-like cognition, enabling multimodal large language models to provide continuous and adaptive service over time.

Fused Deposition Modeling (FDM) is a widely used additive manufacturing (AM) technique valued for its flexibility and cost-efficiency, with applications in a variety of industries including healthcare and aerospace. Recent developments have made affordable FDM machines accessible and encouraged adoption among diverse users. However, the design, planning, and production process in FDM require specialized interdisciplinary knowledge. Managing the complex parameters and resolving print defects in FDM remain challenging. These technical complexities form the most critical barrier preventing individuals without technical backgrounds and even professional engineers without training in other domains from participating in AM design and manufacturing. Large Language Models (LLMs), with their advanced capabilities in text and code processing, offer the potential for addressing these challenges in FDM. However, existing research on LLM applications in this field is limited, typically focusing on specific use cases without providing comprehensive evaluations across multiple models and tasks. To this end, we introduce FDM-Bench, a benchmark dataset designed to evaluate LLMs on FDM-specific tasks. FDM-Bench enables a thorough assessment by including user queries across various experience levels and G-code samples that represent a range of anomalies. We evaluate two closed-source models (GPT-4o and Claude 3.5 Sonnet) and two open-source models (Llama-3.1-70B and Llama-3.1-405B) on FDM-Bench. A panel of FDM experts assess the models' responses to user queries in detail. Results indicate that closed-source models generally outperform open-source models in G-code anomaly detection, whereas Llama-3.1-405B demonstrates a slight advantage over other models in responding to user queries. These findings underscore FDM-Bench's potential as a foundational tool for advancing research on LLM capabilities in FDM.

Predicting roll call votes through modeling political actors has emerged as a focus in quantitative political science and computer science. Widely used embedding-based methods generate vectors for legislators from diverse data sets to predict legislative behaviors. However, these methods often contend with challenges such as the need for manually predefined features, reliance on extensive training data, and a lack of interpretability. Achieving more interpretable predictions under flexible conditions remains an unresolved issue. This paper introduces the Political Actor Agent (PAA), a novel agent-based framework that utilizes Large Language Models to overcome these limitations. By employing role-playing architectures and simulating legislative system, PAA provides a scalable and interpretable paradigm for predicting roll-call votes. Our approach not only enhances the accuracy of predictions but also offers multi-view, human-understandable decision reasoning, providing new insights into political actor behaviors. We conducted comprehensive experiments using voting records from the 117-118th U.S. House of Representatives, validating the superior performance and interpretability of PAA. This study not only demonstrates PAA's effectiveness but also its potential in political science research.